DOI: 10.1590/1414-462X201600010x01

Original Article

Agreement assessment between hemoglobin and hematocrit to detect anemia prevalence in children less than 5 years old Avaliação da concordância entre hemoglobina e hematócrito para detectar prevalência de anemia em crianças menores de 5 anos

Gisela Soares Brunken1, Giovanny Vinícius Araújo de França2,3, Ronir Raggio Luiz4, Sophia Cornbluth Szarfarc5

Abstract This study aimed to assess the extent of agreement between hemoglobin (Hb) and hematocrit (Ht), while also verifying whether anemia prevalence can be determined by hematocrit cut-off points. A hundred pairs of capillary Hb and Ht of children less than 5 years old were analyzed. The Hb and Ht concentrations were measured using HemoCue and microhematocrit centrifuge, respectively. The Bland and Altman’s method was used to assess the agreement between ‘observed Hb’ and ‘Hb estimated through Ht’ (Hb = Ht/3). The ROC curve was constructed based on the reference to anemia classification according to Hb. The subjects were classified according to the status of anemia obtained through the Ht, also evaluating the agreement regarding Hb. The area under the ROC curve identified that Ht <35% correctly classified 85% of subjects with Hb <11.0 g/dL. However, the cut-off point of <33% showed higher kappa (k = 0.49 against 0.41) and higher relative agreement (82% vs. 74%) compared to Ht <35%. Our findings indicate that capillary Hb and Ht could be used either together or separately to assess the prevalence of anemia. Considering the need to identify iron deficiency anemia with a single indicator in population studies, the use of Ht would be simpler and cost-saving. Keywords: hematocrit; hemoglobin; reproducibility of results; child preschool; anemia, iron-deficiency.

Resumo O objetivo deste estudo foi avaliar o grau de concordância entre a hemoglobina (Hb) e o hematócrito (Ht), para verificar se a prevalência de anemia pode ser determinada pelo hematócrito. Cem pares de Hb e Ht capilar de crianças menores de 5 anos de idade foram analisados. A concentração de Ht e Hb foi medida utilizando centrífuga para microhematócrito e HemoCue, respectivamente. A análise de Bland e Altman foi utilizada para avaliar as diferenças entre os valores observados de Hb e de Hb estimada através de hematócrito (Hb = Ht / 3). A curva ROC foi calculada com referência à classificação de anemia ferropriva, a partir dos resultados de Hb. Os indivíduos foram classificados de acordo com o estado de anemia a partir do Ht, comparando com a avaliação obtida a partir da Hb. A área sob a curva ROC permite identificar que Ht <35% corretamente classificava 85% dos indivíduos com Hb <11,0 g/dL. No entanto, o ponto de corte Ht <33% apresentou maior kappa (k = 0,49 contra 0,41) e concordância relativa mais elevada (82% vs. 74%) em comparação com Ht <35%. A hemoglobina e o hematócrito capilar podem ser usados em conjunto ou separadamente para avaliar a prevalência de anemia. Considerando a necessidade de identificar a anemia ferropriva com um único indicador em estudos populacionais, a utilização do Ht seria mais simples e econômica. Palavras-chave: hematócrito; hemoglobina; reprodutibilidade dos testes; pré-escolar; anemia ferropriva.

1Departamento de Saúde Coletiva, Universidade Federal de Mato Grosso (UFMT) - Cuiabá (MT), Brazil. 2Centro de Pesquisas Epidemiológicas, Universidade Federal de Pelotas (UFPEL) - Pelotas (RS), Brazil. 3MRC Epidemiology Unit, Institute of Metabolic Science, Addenbrooke’s Hospital - Cambridge, UK. 4Instituto de Estudos em Saúde Coletiva, Universidade Federal do Rio de Janeiro (UFRJ) - Rio de Janeiro (RJ), Brazil. 5Faculdade de Saúde Pública, Universidade de São Paulo (USP) - São Paulo (SP), Brazil. Study carried out at Federal University of Mato Grosso (UFMT) – Cuiabá (MT), Brazil. Correspondence: Gisela Soares Brunken – Rua Buenos Aires, 726, apto 102, CEP: 78060-634 – Cuiabá (MT), Brazil – Email: [email protected] Financial support: none. Conflict of interests: nothing to declare.

118 Cad. Saúde Colet., 2016, Rio de Janeiro, 24 (1): 118-123 Hb and ht in field work

▄▄INTRODUCTION practical equipment for use in field work. It requires a simple Iron deficiency anemia is the world’s most prevalent and short training period to operate, and a stable power source. nutritional deficiency, affecting several countries worldwide, It requires about 0.5 mL of blood, equivalent to a drop, extracted with serious consequences for human health, social development in a capillary tube, closed at one end and centrifuged6,7. As in and economy1. Its etiology is diverse and can be attributed to portable hemoglobinometers (HemoCue and Agabê), it does not require transportation of samples and the results are released nutritional deficiencies of iron, folic acid, vitamin A, C or B12, as well as numerous diseases, such as malaria, intestinal parasites, in a few minutes. The same method is used by blood banks to hemoglobinopathy. Iron deficiency has been identified as the identify donors feasibility. One limitation is the need of a high main contributing factor to anemia, so that the terms “anemia” centrifugal force to prevent a significant amount of plasma being and “iron deficiency” are often treated as synonyms. As iron trapped between the red blood cells. This can be eliminated by deficiency anemia is the most severe stage of the iron deficiency, the use of a stable power source7. it is estimated that there are about 2.5 times more individuals The hematocrit below 33% is an acceptable indicator for that are iron deficient than anemic2,3. Studies have shown that evaluation and identification of anemic patients with hemoglobin anemia persists over last decades at high levels, during infancy in place2. There is a standard conversion between the two 4,5 and childhood . measures (Hb = Ht/3) which is commonly used to define cut-off 9 Different from anemia driven by vitamin B12 or folic acid points for estimating the prevalence of anemia . Health services deficiency, iron deficiency anemia is microcytic and hypochromic. has been using this indicator for anemia screening, since it is There are several well-described laboratory measurements to access less expensive and more practical, but reliability studies are and monitor the nutritional status of iron, but only hemoglobin needed. In this study, we aimed to assess the agreement between 6,7 and hematocrit are routinely performed in field trials . hemoglobin and hematocrit capillary under field conditions in The hemoglobin concentration lower than 11 g/dL indicates children less than 5 years old, in order to identify a cut-off point anemia, regardless the etiology2. Venous blood measurement with adequate psychometric characteristics for the outcome is well standardized6,7, however practical issues limit its use in of hematocrit sensitive to identify individuals with microcytic epidemiological studies. Therefore, Hb concentration is commonly anemia. It is important to investigate the advantage of using the measured in capillary blood, using an accurate point‑of-care test hematocrit instead of hemoglobin. In fact, it is not to say that such as the HemoCue (HemoCue AB, Angelholm, Sweden)8. one method is better than another, but knowing that the two Collecting a drop of blood using a finger stick, rather than are consistent and, judging by the hematocrit, has its validity. collecting venous blood using a syringe, is a significant advantage in the field research. Moreover, it is fast, portable, shows directly ▄▄MATERIALS AND METHODS the WHO anemia indicator and it is safer because it does not require aspiration or syringe. Unfortunately the cost of using This is a cross sectional study carried out in 2008 including HemoCue is high when compared to laboratory testsa and, once 100 children less than 5 years old who attended three public the seal is removed, the tubes with 50 micro cuvettes have a shelf day care centers in the city of Cuiabá, a city located in the life of only three months. The Agabê deviceb is also available Center‑West region of Brazil, state of Mato Grosso, with an since 2009, which is developed and produced in Brazil (M-Exa estimated population of 569.830 inhabitants. A previous study Biomedical Instrumentation). It allows dosing hemoglobin in showed high prevalence (63%) of anemia among children aged a much more cost efficient manner. As the HemoCue device, less than 36 months in public day care centers of Cuiabá10. it informs the hemoglobin concentration in capillary sample Weight was measured on an electronic scale (capacity: in real time, but requires dilution of the blood drop on the 150 kg; precision: 50 g) and height using an anthropometric reagent, demanding cautious for treatment and evaluation of scale (precision: 0.5 cm). Nutritional status was classified as: the sample in the field. underweight – weight for age <–2 standard deviations (SD) of The hematocrit (or packed cell volume) is the percentage of the WHO Child Growth Standards median11 –; stunting – height volume occupied by red blood cells compared to the total blood for age <–2 SD – and overweight – weight-for-height >+2 SD. volume. It is obtained using a centrifuge microhaematocrit, The associations between presence of anemia with sex, age and c a simple and cost-saving approach , which is a portable and nutritional status were investigated using the qui-squared test. Samples of capillary hematocrit and hemoglobin were collected At July 2014. using the methodology recommended by the manual of the a HemoCue = R$ 4,850.00; Case with 200 microcuvettes = R$ 880.00 HemoCue equipment4. Participants were classified according to (R$ 4.40 / microcuvette). the status of anemia from hemoglobin concentrations, according b Agabê = R$ 120.00. 2 c Centrifug for microhematocrit, capacity for 24 capillary tubes, to the cut-off point recommended by WHO . Regarding the Centribio = R$ 2,900.00. cut-off points, we applied international recommendations on

Cad. Saúde Colet., 2016, Rio de Janeiro, 24 (1): 118-123 119 Gisela Soares Brunke, Giovanny Vinícius Araújo de França, Ronir Raggio Luiz, Sophia Cornbluth Szarfarc

the diagnosis and classification of anemia2, classifying as normal Analyses were performed using Stata, version 11.0. The study hemoglobin values lower than 11.0 g/dL and hematocrit lower protocol was approved by the Ethics Committee in Research than 33%. The prevalence of anemia was estimated according of Hospital Universitário Julio Muller, Universidade Federal to nutritional status and age. do Mato Grosso. The Bland and Altman method12 was used to assess the ▄ agreement between the two estimates of anemia status – ▄ RESULTS ‘hemoglobin observed’ and ‘estimated hemoglobin’ (Hb = Ht/3), The average age was 35 months (range 13-56) among the estimating the 95% limits of agreement from the differences 100 children evaluated (53 girls), and the mean weight and observed. The Receiver Operating Characteristic curve (ROC) height were similar to their median: 13.5 kg (median 13.6) and was constructed in order to identify the hematocrit value that 91.6 cm (median 91.8). The mean concentrations of hemoglobin maximizes the sensitivity and specificity in screening for anemia, and hematocrit were 11.7 g/dL (median 11.9) and 36.2% with reference to the classification obtained from the capillary (median 36.0), respectively (Table 1). The presence of anemia hemoglobin, comparing the ‘crude’ and ‘adjusted for age and sex’ was significantly associated only with age Table 2( ). ROC area. Subjects were also classified according to the status The Bland and Altman plot Figure 1( ) presents the differences of anemia from hematocrit, using the cut-off point proposed by between observed and estimated hemoglobin from hematocrit. WHO2 and identified from the ROC curve by calculating the The differences were inversely associated with the mean of kappa coefficient to verify the degree of agreement in relation measured and estimated Hb (r = –0.445, p <0.001), and the to the classification by hemoglobin. 95% limits of agreement were –3.645 to 2.959 g/dL.

Table 1. Sample description of 100 children less than 5 years old assessed in public nurseries. Cuiabá – Mato Grosso, 2008 Descriptive Statistics Variables Standard N Mean Median Range Deviation Age (months) 91 35.1 11.3 34.2 13-56 Weight for Height (scores z) 100 0.23 0.91 0.20 –2.46-2.44 Height for Age (scores z) 91 –0.53 1.17 –0.47 –4.45-4.66 Weight for Age (scores z) 91 –0.13 0.94 –0.16 –2.62-2.68 BMI for Age (scores z) 91 0.28 0.96 0.23 –2.61-2.6 Hemoglobin (g/dl) 100 11.7 1.2 11.9 7.9-14.7 Hematocrit (%) 100 36.2 5.5 36.0 19.0-55.7

Table 2. Prevalence of anemia of 100 children less than 5 years old assessed in public nurseries, according to sex, age and nutritional status. Cuiabá – Mato Grosso, 2008 Prevalence of anaemia* Variable p-value** N % CI 95% Overall (n=100) 23 23.0 15.7-32.4 - Sex (n=100) Male 13 27.7 10.3-32.0 0.297 Female 10 18.9 16.6-42.4 Age (n=91) <24 months 8 47.1 24.7-70.6 0.002 ≥24 months 10 13.5 7.3-23.6 Underweight (n=91) No 16 18.4 11.5-28.1 0.121 Yes 2 50.0 9.2-90.8 Stunting (n=91) No 16 19.1 11.9-29.1 0.543 Yes 2 28.6 6.2-70.7 Overweight (n=100) No 21 21.9 14.6-31.4 0.190 Yes 2 50.0 9.2-90.8 *Hb<11g/dL **Pearson’s chi-squared test

120 Cad. Saúde Colet., 2016, Rio de Janeiro, 24 (1): 118-123 Hb and ht in field work

The area under the ROC curve was high (85%) and significant (specificity) Table 3( ). There were no differences among crude (p <0.001), so the hematocrit value below 35% seems to correctly and adjusted curves for age and sex (p>0.05). identify 85% of hemoglobin below 11.0 g/dl (Figure 2). This cut-off Table 3 compares the sensitivity and specificity in predicting point (35%) presents a sensitivity of 78.3% to identify children hematocrit anemia according to the two cut-off points: Ht <33%2 with Hb <11.0 (actually anemic) and 72.5% of non-anemic and Ht <35% (this study). The specificity was higher (88.3%

Figure 1. Bland and Altman graph about the difference among hemoglobin measured and estimated from the hematocrit (Hb = Ht/3), by their average values in children less than 5 years old. Cuiabá – Mato Grosso, 2008. Diferences among crude and ajusted curves (X2 = 1.75; Prob>X2 = 0.1864)

Figure 2. ROC curve of Hematocrit and diagnosis of anemia in children less than 5 years old, adjusted for age and sex. Cuiabá – Mato Grosso, 2008

Cad. Saúde Colet., 2016, Rio de Janeiro, 24 (1): 118-123 121 Gisela Soares Brunke, Giovanny Vinícius Araújo de França, Ronir Raggio Luiz, Sophia Cornbluth Szarfarc

Table 3. Psychometric characteristics and agreement on the classification of anemia with hematocrit compared to that obtained by capillary hemoglobin in children less than 5 years old. Cuiabá – Mato Grosso, 2008 Anemia by hemoglobin Relative Sensibility Specificity Anemia by Hematocrit (Hb<11g/dl) Kappa Concordance (%) (%) Anemic Not anemic (%) Ht<33% Anemic 14 9 60.9 88.3 0.49 82.0 Not anemic 9 68 Ht<35% Anemic 18 21 78.3 72.5 0.41 74.0 Not anemic 5 56 r = –0.445, p <0.001

versus 72.5%) for the cut-off point <33%, but the sensitivity Other researchers have used the hematocrit to estimate was lower (60.9% versus 78.3). The cut-off point <33% showed the prevalence of anemia in malaria endemic areas, reporting a better but moderate kappa coefficient (k = 0.49 against that the hematocrit underestimates the prevalence of anemia 0.41)13. The relative agreement between the estimate of anemia assessed by hemoglobin16,17. Graitcer et al.18 studied more than (Hb <11.0 g/dl) using a cut-off of hematocrit <33% was also 10,000 pairs of examinations in children under 2 years of age better (82% vs. 74%) when compared to the cut-off of <35%. and concluded that hemoglobin usually detects higher rates of anemia than hematocrit, as they identified more false negatives ▄▄DISCUSSION using the hematocrit <33%. The suggestion of this study to increase the cut-off of hematocrit <33% to <35% is one more The cut-off for hematocrit <33% had higher relative argument to better assess its prevalence. agreement and kappa coefficient in relation to the concentration Quintó et al.19, studying paired values from almost of hemoglobin. It is important to remember that the relative 3,500 children under 5 years of age living in malaria-endemic agreement and kappa evaluate repeatability, but not necessarily areas in sub-Saharan Africa, observed that the Hb/Ht ratio validity. In addition, the kappa is strongly influenced by was not exactly 1/3 and it varies during the first years of life. 14 prevalence, as well as by positive and negative predictive values . They also pointed out that the commonly assumed ‘equivalent’ To assess validity, psychometric measurements – sensitivity and cut-off points for anemia definition need to be re-evaluated. specificity – are more appropriate, verifying the best checking Flores-Torres et al.20, studying 6,004 individuals, observed an point that maximizes both15. Therefore, it is reasonable to use a overestimation of hemoglobin levels estimated from hematocrit cut-off <35% due to the need for attracting more individuals at levels and, therefore, an underestimation of the prevalence risk of deficiency. In addition, we found little variation on the of anemia. This is another argument for adopting a higher relative agreement, less than 10%, as well as a moderate kappa cut‑off point (<35%) to estimate the presence of anemia from between the two cut-off points13. the hematocrit levels. Increasing the cut-off point from 33% Considering the Bland and Altman graphic (Figure 1), to 35% would allow identifying more individuals at risk of the bias seems to occur between proportional values close iron deficiency. to the mean hemoglobin: 10 and 14 g/dL. That is, in a range The Brazilian Ministry of Health has implemented several programs to prevent and treat iron deficiency anemia. It is close to the central values probably it will not occur medium crucial to have fast and cost-saving approaches to estimate estimation errors, in the case of using hematocrit to estimate the prevalence of anemia, especially in Brazil where universal the concentration of hemoglobin. intervention are being used, such as supplementation and This study small sample size has limited power to predict the fortification of foods with iron21. hemoglobin from hematocrit, but is sufficient to investigate the correlation between these two capillary measures. Surely the ▄▄CONCLUSION ideal is always to use more than two capillary measures, for Hemoglobin and hematocrit capillary can be used together better assessment of the nutritional status, as measured by or separately to track and assess the prevalence of anemia. the hematocrit and hemoglobin hypochromic microcytosis. In Brazil, where the prevalence of iron deficiency anemia in However in field work, and also to track individuals and to have children reaches epidemic proportions, Ht could be easily assessed a rapid diagnosis, the health manager and the researcher often in different settings, providing estimates of the prevalence of have to choose only one measure or instrument. In this case, anemia and helping designing and implementing strategies to the hematocrit seems to be a good option even if used alone. prevent and treat anemia.

122 Cad. Saúde Colet., 2016, Rio de Janeiro, 24 (1): 118-123 Hb and ht in field work

▄▄REFERENCES

1. World Health Organization. Worldwide prevalence of anemia 1993-2005: 13. Altman DG. Practical statistics for medical research. London: Chapman WHO global database on anemia. Geneve; 2008. & Hill; 1991.

2. World Health Organization. Iron deficiency anemia: assessment, prevention 14. Hoehler FK. Bias and prevalence effects on kappa viewed in terms of and control. Geneve: UNICEF/UNU; 2001. sensitivity and sensitivity and specificity. J Clin Epidemiol. 2000;53(5):499- 503. http://dx.doi.org/10.1016/S0895-4356(99)00174-2. PMid:10812322. 3. Batista Fo M, Souza AI, Bresani CC. Anemia como problema de saúde pública: uma realidade atual. Cien Saude Colet. 2008;13(6):1917-22. http:// 15. Fletcher RH, Fletcher SW. Epidemiologia clínica: elementos essenciais. São dx.doi.org/10.1590/S1413-81232008000600027. PMid:18833369. Paulo: Art Med; 2006. 188 p.

4. Kassebaum NJ, Jasrasaria R, Naghavi M, Wulf SK, Johns N, Lozano R, et al. 16. Carneiro IA, Drakeley CJ, Owusu-Agyei S, Mmbando B, Chandramohan A systematic analysis of global anemia burden from 1990 to 2010. Blood. D. Haemoglobin and haematocrit: is the threefold conversion valid for 2014;123(5):615-24. http://dx.doi.org/10.1182/blood-2013-06-508325. assessing anaemia in malaria-endemic settings? Malar J. 2007;6(1):67-71. PMid:24297872. http://dx.doi.org/10.1186/1475-2875-6-67. PMid:17519019.

5. Granado FS, Augusto RA, Muniz PT, Cardoso MA. Anaemia and iron 17. Rodríguez-Morales AJ, Sánchez E, Arria M, Vargas M, Piccolo C, Colina deficiency between 2003 and 2007 in Amazonian children under 2 years of R, et al. Haemoglobin and haematocrit: the threefold conversion is also age: trends and associated factors. Public Health Nutr. 2013;16(10):1751-9. non valid for assessing anaemia in Plasmodium vivax malaria-endemic http://dx.doi.org/10.1017/S1368980012005617. PMid:23369207. settings. Malar J. 2007;6(1):166-9. http://dx.doi.org/10.1186/1475-2875- 6-166. PMid:18086306. 6. International Nutritional Anemia Consultative Group. Measurements of iron status. Washington: ILSI Press; 1985. 18. Graitcer PL, Goldsby JB, Nichaman MZ. Hemoglobins and hematocrits: are they equally sensitive in detecting anemias? Am J Clin Nutr. 1981;34(1):61- 7. World Health Organization. Assessing the iron status of populations. 4. PMid:7446459. Geneve; 2007. 19. Quintó L, Aponte JJ, Menéndez C, Sacarlal J, Aide P, Espasa M, et al. 8. von Schenck H, Falkensson M, Lundberg B. Evaluation of “HemoCue”, a Relationship between haemoglobin and haematocrit in the definition new device for determining haemoglobin. Clin Chem. 1986;32(3):526-9. of anaemia. Trop Med Int Health. 2006;11(8):1295-302. http://dx.doi. PMid:3948400. org/10.1111/j.1365-3156.2006.01679.x. PMid:16903892.

9. Bain BJ, Bates I. Basic haematological techniques. In: Lewis SM, Bain BJ, 20. Flores-Torres J, Echeverría-Ortega M, Arria-Bohorquez M, Hidalgo G, Bates I. Practical hematology. 9th ed. Edinburgh: Churchill Livingstone; Albano-Ramos C, Sanz R, et al. Diferencias entre la hemoglobina observada 2001. p. 19-46. y estimada por hematocrito y su importancia en el diagnóstico de anemia en población costera Venezolana: análisis del segundo estudio nacional de crecimiento y desarrollo humano (SENACREDH). Rev Peru Med 10. Brunken GS, Guimarães LV, Fisberg M. Anemia em crianças menores de Exp Salud Publica. 2011;28(1):47-53. http://dx.doi.org/10.1590/S1726- 3 anos que freqüentam creches públicas em período integral. J. Pediatr. 46342011000100008. PMid:21537769. 2002; 78(1):50-6. http://dx.doi.org/10.2223/JPED.664. 21. Szarfarc SC. Políticas públicas para o controle da anemia ferropriva. Rev 11. World Health Organization. Nutritional Landscape Information System Bras Hematol Hemoter. 2010;32(Suppl 2):2-7. http://dx.doi.org/10.1590/ (NLIS) country profile indicators: interpretation guide. Geneve; 2010. S1516-84842010005000065.

12. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurements. Lancet. 1986;1(8476):307-10. http:// Received on: Oct. 09, 2015 dx.doi.org/10.1016/S0140-6736(86)90837-8. PMid:2868172. Accepted on: Feb. 01, 2016

Cad. Saúde Colet., 2016, Rio de Janeiro, 24 (1): 118-123 123